Some simply ploughed on regardless not changing a thing. Some reduced their asks for a couple of years and then returned to their earlier, higher ask – presumably because the change didn’t work.

Others raised their asks for a considerable time and then reduced them again – presumably that didn’t work either. Yet others were simply all over the place.

With a bit of disciplined planning it is actually very easy to discover quite quickly and definitively the combination of asks which works best at any given time. The answer is, of course, the tried and tested A & B split run. Take the Sightsavers examples shown here (no idea if they were split-run tested or not), so long as they can estimate the lifetime value of a donor and know that they get, on average, x per cent of their total response to a newspaper ad within five days of publication, they can make a pretty good stab at which half of a split run has ‘won’ within a week. 

This speed of analysis is vital for three reasons: first it saves you from running too long with an unsuccessful ad; second it means that, once you’ve tested a successful ad, you need waste no time in rolling it out to take advantage of its enhanced performance and third, very importantly, it can give you early warning when your banker ad inevitably hits the ‘wear-out’ wall and allow you to pull it immediately to avoid a disaster.

Sadly, too many charities do their response analysis on a campaign basis without split runs – leaving it to the end of the campaign to see which copy or which ask variant has come out top. This can be disastrously misleading, not only for the reverse of the reasons mentioned above but because the response to individual inserts can be influenced by so many extraneous factors that it renders the comparisons worse than useless – even downright damaging.

As invaluable as A & B split-runs are, they are not without their dangers. For absolutely reliable results it is imperative that only one element of an ad is tested at a time. The temptation to kill two birds with one stone must be resisted because it will only cloud the issue. The Sightsavers ads above, for example, would not make an ideal split-run test because it would be not just testing TEXT v. coupon but one means of response (TEXT) v. two (FREEPOST coupon and 0800 telephone number) and because the TEXT version says £3 will treat three children but the couponed versions says £17.50 will treat 50 children. Apart from the fact that the sums don’t compute, it would be impossible to separate the elements causing one version to perform better than the other.

Because newspapers typically charge 10 per cent extra for a split-run, such testing can seem an expensive luxury but it can pay for itself many times over – and a good buyer may even be able to cut a deal. A very cost-effective solution can be to use loose inserts, which can easily be printed – on a long enough run – with as many as five tests interleaved in order to try five separate elements or copy variants. Loose inserts can be more cost-effective than ads anyway and give you much greater creative scope.

Even small changes can, and should be, tested, such as reversing the order of a range of asks from highest-to-lowest, adding examples of what the money might buy, changing the number of steps in the range of asks or the distances between each step, etc.

I recommend the high assay approach to off-the-page fundraising. This involves putting the bulk of your budget behind your proven banker ad but setting aside, say, around 20 per cent for a carefully planned programme of testing using split-runs – to cover both new creative strategies and the nuts and bolts of ways of improving the performance of the banker. It is a fool-proof way of optimising performance, insuring against unexpected disasters and minimising the deleterious impact of political, economic, competitive and social changes outside your control.

If you’re not using split-runs systematically, you should be. There is no alternative.